Skip to main content
NCBI home page
Neurology logoLink to Neurology
. 2021 Jan 19;96(3):e366–e375. doi: 10.1212/WNL.0000000000011109

Contribution of Ictal Source Imaging for Localizing Seizure Onset Zone in Patients With Focal Epilepsy

Shuai Ye 1,*, Lin Yang 1,*, Yunfeng Lu 1, Michal T Kucewicz 1, Benjamin Brinkmann 1, Cindy Nelson 1, Abbas Sohrabpour 1, Gregory A Worrell 1, Bin He 1,✉
PMCID: PMC7884986  PMID: 33097598

Abstract

Objective

To determine whether seizure onset zone (SOZ) can be localized accurately prior to surgical planning in patients with focal epilepsy, we performed noninvasive EEG recordings and source localization analyses on 39 patients.

Methods

In 39 patients with focal epilepsy, we recorded and extracted 138 seizures and 1,325 interictal epileptic discharges using high-density EEG. We investigated a novel approach for directly imaging sources of seizures and interictal spikes from high-density EEG recordings, and rigorously validated it for noninvasive localization of SOZ determined from intracranial EEG findings and surgical resection volume. Conventional source imaging analyses were also performed for comparison.

Results

Ictal source imaging showed a concordance rate of 95% when compared to intracranial EEG or resection results. The average distance from estimation to seizure onset (intracranial) electrodes is 1.35 cm in patients with concordant results, and 0.74 cm to surgical resection boundary in patients with successful surgery. About 41% of the patients were found to have multiple types of interictal activities; coincidentally, a lower concordance rate and a significantly worse performance in localizing SOZ were observed in these patients.

Conclusion

Noninvasive ictal source imaging with high-density EEG recording can provide highly concordant results with clinical decisions obtained by invasive monitoring or confirmed by resective surgery. By means of direct seizure imaging using high-density scalp EEG recordings, the added value of ictal source imaging is particularly high in patients with complex interictal activity patterns, who may represent the most challenging cases with poor prognosis.

Surgical treatment provides an effective curative solution for focal epilepsy.1 To achieve seizure freedom, the epileptogenic zone (EZ) has to be removed or disconnected.2–4 Thus, a comprehensive presurgical evaluation to delineate the EZ is crucial for a satisfactory outcome.

Since the 1930s, EEG has been an essential component of the presurgical evaluation due to its high temporal resolution and its capability of long-term recordings.5–7 The limited spatial resolution of scalp EEG has also been essentially addressed by high-density EEG recording systems and source imaging techniques.8–10 In the past decade, EEG source imaging has emerged as a viable option for guiding surgical interventions.8–11

Many studies have used signals recorded during the interictal period8,12 instead of the ictal period,13–16 as the recording, interpretation, and analysis of seizure activity remains a challenge. While the convenience of recording interictal activities must be acknowledged, whether the irritative zone (IZ) could serve as a good indicator of the EZ is debatable.3,11 Moreover, interictal activities are known to propagate through the cortex17 and occur contralateral to seizure onset,18–20 which has posed great challenges to clinical diagnosis.

We have investigated the performance of a long-term dense-array EEG monitoring protocol and source imaging technology for both ictal and interictal recordings to localize seizure onset zone (SOZ). In a group of patients with focal epilepsy, we sought to determine quantitatively whether (1) noninvasive seizure imaging is concordant with the clinical diagnosis and (2) ictal recordings could provide additional information compared to interictal recordings.

Materials

Patient Enrollment and Data Acquisition

We studied patients who underwent presurgical evaluation with high-density long-term EEG monitoring at Mayo Clinic, Rochester, Minnesota, from 2007 to 2017. Patients with epilepsy were consecutively enrolled in low-density EEG studies first, but were asked to participate in the high-density EEG study if the initial low-density EEG evaluation indicated that the patient may need further investigation. Patients were included in the study if they underwent at least one invasive procedure: (1) surgical resection and a postoperative follow-up of at least 12 months; (2) long-term invasive intracranial EEG (iEEG) monitoring. Surgical outcomes were defined based on the International League Against Epilepsy (ILAE) scale.21

Each patient underwent long-term video EEG monitoring using a commercial EEG system (XLTEK, Natus Medical Incorporated). A total of 76 individual electrodes were attached to the scalp according to the 10-10 system with a typical interelectrode distance of approximately 3 cm. The EEG signals were recorded with a 500-Hz sampling rate. The dense-array EEG was continuously recorded for at least 24 hours (and may have lasted for 6 days). All patients underwent an anatomical MRI scan as part of their presurgical evaluation. An individual 3-layer boundary element head model was constructed from MRI, and anatomical landmarks were identified to coregister a generic electrode position montage to the head shape of each individual. Anatomical MRI scans obtained 3–4 months postoperatively were available in some of the patients who underwent surgical resection and were used to determine the boundaries of the surgical resection. The subset of patients who underwent iEEG monitoring had CT images.

In the patients who subsequently underwent iEEG monitoring to determine the SOZ, the iEEG waveforms were carefully reviewed by board-certified epileptologists. The iEEG electrodes involved in the seizure onset were determined by visual inspection. The locations of the iEEG electrodes were segmented from CT images and coregistered with the patient's MRI (iEEG-SOZ). In patients with surgical resection, resected regions were determined from the postoperative MRI and coregistered to preoperative MRI. Some patients who underwent a standard temporal lobectomy did not undergo a postoperative MRI. In these patients, the surgically removed regions were determined according to the surgery notes. An illustrative analysis pipeline can be found in figure 1.

Figure 1. Illustrative Analysis Pipeline.

Figure 1

Open in a new tab

The data set we obtained for each patient includes preoperational MRI, high-density long-time scalp EEG recordings with notes from experienced epileptologists, intracranial EEG (iEEG) recordings, CT scans containing iEEG electrode positions (if available), postoperational MRI (if available), and follow-up information. Yellow blocks indicate the information extracted from the raw data, such as the individual boundary element method (BEM) models. Blue blocks indicate the analysis we have done. Top right panel shows the obtained analysis outcome (irritative zone from interictal source imaging and EEG seizure onset zone [SOZ] from ictal source imaging). Bottom right panel shows how we extract clinical evaluation made by the clinicians from a consensus of various information (i.e., iEEG SOZ, and resected volume determined from EEG, MRI, SPECT, semiology). ILAE = International League Against Epilepsy.

Standard Protocol Approvals, Registrations, and Patient Consents

The study was approved by the institutional review board of the Mayo Clinic (Rochester, MN), University of Minnesota (Minneapolis), and Carnegie Mellon University (Pittsburgh, PA). All patients gave written informed consent before entering the study.

Ictal Analysis

For each patient, the EEG waveforms were visually reviewed by board-certified epileptologists to identify onset of seizures. For each seizure, we segmented the ictal epoch with a short preictal period (∼30 s). The ictal EEG epoch was analyzed using the dynamic seizure imaging (DSI) technique.22 From the spatiotemporal cortex activation results of DSI, the spectral power in the frequency range of interest (ictal rhythms) was calculated for each voxel in the source space in a 1-second sliding window with 0.5-second overlap. The SOZ of the estimated source was computed as the spectral power distribution in the interval at seizure onset. We then compared the DSI results using different number of scalp electrodes. Scalp recordings were down-sampled from 76 channels to 32 and 19 channels as is routine in clinical settings, and SOZ for each configuration was estimated using the same DSI algorithm. Localization error was computed and compared in different sensor configurations. The down-sampling occurred when solving the source imaging problem. We still used the full EEG information for independent component analysis.

We also performed ictal analysis with another commonly used frequency domain–based method.23,24 In this frequency domain–based imaging (FDI) method, seizure segment at the onset was transformed into Fourier space. The source distribution of the dominant frequency was then constructed. Details regarding the 2 methods are available from Dryad (appendix) at doi.org/10.5061/dryad.4b8gtht9n.

Interictal Analysis

For each patient, the scalp EEG data were bandpass filtered between 1–50 Hz and bad channels were removed offline, i.e., during preprocessing, and not when recording EEG. The data were then visually inspected and marked for interictal epileptic discharges (IEDs) using Curry software (Compumedics, Charlotte, NC) by 2 experienced researchers, independently. IED included spikes (interval width of 20–70 ms) and sharp waves (interval width of 70–200 ms). Physiologic artifacts or rhythms were rejected. Only IEDs without co-occurring artifacts (muscle, eye movements, blinks, or electrocardiographic artifacts) were selected for further evaluation.

The extracted IEDs were classified according to duration, morphology, and topology maps at IED peaks. At least 5 IEDs were required to form a cluster. If more than 90% of the IEDs were in the same cluster, then only the dominant cluster was included. Otherwise all the formed clusters were included for subsequent analysis. After obtaining clustered IEDs for all patients, IEDs in the same cluster were aligned to the peak and averaged, to increase the signal-to-noise ratio (SNR). The sLORETA algorithm25 was applied on the peak of the averaged IED for each IED cluster. Two researchers had to agree on the classification and localization of the interictal findings. If more than one cluster of IEDs were found in a patient, the IZs estimated from each cluster were included in the analysis.

Evaluation Metrics

Qualitative and quantitative assessments were used to evaluate results. The qualitative assessment of the analysis is the sublobar concordance rate of estimated EEG-SOZ with clinical findings. Basically, the cortex was segmented into 20 regions (10 for each hemisphere) based on anatomical landmarks such as the central sulcus, the sylvian fissure, et cetera, as reported in literature,26,27 and the outcome was defined as concordant if the estimated source maximum fell within the same cortical region as the resection/iEEG SOZ. The primary quantitative outcome of the analysis is localization error (LE). LE was calculated as the Euclidean distance from the estimated source maximum and the closest iEEG-SOZ electrodes if iEEG study was available, or as the distance from the estimated source to the boundary of the resected volume if postoperative MRI was available.

Statistical Analysis

For ictal EEG analysis, we assessed the effect of different montages. Welch t test was applied on the LEs between different montages (i.e., 76, 32, and 19 channels), to determine whether number of recoding channels has any effect on the source imaging performance. For interictal EEG analysis, since each patient might have had more than one IED cluster, LEs were regrouped into 3 categories for each patient: all clusters (including LE of all IED clusters), major cluster (including the IED cluster with maximum number of IEDs), and closest cluster (including the IED cluster with smallest LE). Two-tailed Wilcoxon rank-sum test was used to compare the LE of ictal and interictal EEG analysis outcomes.

For all statistical analyses, a p value of less than 0.05 was considered to indicate statistical significance. All the analyses of the current study were performed with the use of R, version 3.4.3, which is a well-known statistical package.

Data Availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Results

Patient Inclusion

We assessed 78 patients enrolled in Mayo Clinical for high-density EEG recording during the period of January 2010–May 2017. A total of 39 adult patients with partial seizures were included in the study according to the study's criteria (outlined in figure 2). A total of 51.28% (20/39) are female. The averaged onset age is 17.87 years, SD 13.41, and the average duration of epilepsy is 15.47 years, SD 12.58 (onset age information is not known for 4 patients). Among these 39 patients, 20 had iEEG recordings available and 28 had postoperational MRI available (in 34 patients who went through surgical resection). After the surgery, 18 patients became seizure-free (ILAE 1–2 outcome) and 10 patients continued to have recurrent seizures (ILAE 3–6 outcome) according to their most recent follow-ups. Patient enrollment and outcomes are summarized in figure 2. A more detailed patient information table can be found in Dryad (appendix) at doi.org/10.5061/dryad.4b8gtht9n.

Figure 2. Patient Inclusion.

Figure 2

Open in a new tab

A total of 39 adult patients were included in the study after applying the criteria on the patient pool. Among the 39 patients, 20 of them had intracranial EEG (iEEG) recording and 28 of them went through surgical resection. After the surgery, 18 patients became seizure-free (International League Against Epilepsy [ILAE] 1–2 outcome) and 10 patients had recurrent seizures (ILAE 3–6 outcome) according to their most recent follow-up. Note that among 20 patients with iEEG recording, surgical outcome of 6 patients were known while their postoperative images were not available. The 6 patients were not included among the 28 patients with surgical resection. ILAE-1 = completely seizure-free outcome; ILAE-2 = only auras, no other seizures; ILAE-3 = 1 to 3 seizure days per year; ILAE-4 = 4 seizure days per year to 50% reduction of baseline seizure days; ILAE-5 = less than 50% reduction of baseline seizure days; ILAE-6 = more than 100% increase of baseline seizure days. E-TL = extratemporal lobe; TL = temporal lobe.

Ictal EEG Analysis Outcome

A total of 138 seizures were included (3.54 per patient, SD 2.53) in this study. The sublobar concordance between EEG-SOZ and clinical evaluation for all patients was 94.87% (37/39). For 20 patients who underwent iEEG recording, the mean LE is about 2.02 cm (SD 2.19 cm, min 0.41 cm, max 8.58 cm). However, DSI identified sources at regions without iEEG coverage in 2 cases where the patients were not seizure-free. Details are available from Dryad (appendix) at doi.org/10.5061/dryad.4b8gtht9n. By removing these 2 cases, the mean LE is approximately 1.35 cm (SD 0.69 cm, min 0.41 cm, max 3.11 cm). For 28 patients who had postoperative MRI, the mean LE between the estimated source maximum and the resection is approximately 1.20 cm (SD 1.64 cm, min 0 cm, max 6.43 cm). However, in the subset of patients who became seizure-free, the LE is much smaller. Among the 28 patients, 18 of them had ILAE 1–2 outcome at the most recent follow-up, which means they became seizure-free for at least 1 year. The DSI results for 18 seizure-free patients who had postoperative MRI show a mean LE of approximately 0.74 cm (SD 0.79 cm, min 0 cm, max 2.45 cm) to the boundary of the resection. Examples of the DSI results can be found in figure 3 and example videos are available from Dryad (movies S1–S4) at doi.org/10.5061/dryad.4b8gtht9n.

Figure 3. Source Imaging Examples.

Figure 3

Open in a new tab

Estimated source at seizure onset in (A) 1 temporal seizure with the reconstructed spectrogram from 2 regions of interest (ROIs) in the source domain; (B) 1 extratemporal lobe seizure with the reconstructed spectrogram from a ROI; and (C) 1 temporal lobe seizure presented with brain tissues removed in the surgery. More examples can be found in movies 1–4 (doi.org/10.5061/dryad.4b8gtht9n). The left panel shows the implantation or surgical resection. The middle panel shows the dynamic seizure imaging results overlapped on the brain. The right panel (for A and B) shows the spectrogram where the x-axis is the time window (second) and the y-axis is the frequency range (Hz).

With the FDI method, in 18 patients with iEEG recording, the mean LE is approximately 2.30 cm (SD 1.71 cm, min 0.49 cm, max 8.33 cm) compared to SOZ. In 18 patients who had postoperative MRI and ILAE 1–2 surgical outcome, the mean LE is approximately 1.08 cm (SD 0.95 cm, min 0 cm, max 3.25 cm). As shown in figure 4, the localization error obtained using FDI method is larger than LE obtained using DSI method on average.

Figure 4. Comparison of Dynamic Seizure Imaging (DSI) Results With Frequency Domain Imaging (FDI).

Figure 4

Open in a new tab

(A) Localization error (LE) in comparison with intracranial EEG (iEEG) in the 18 patients with iEEG recordings excluding 2 outliers. (B) LE in comparison with resection and postoperative MRI in the 18 patients with International League Against Epilepsy 1–2 outcome. The blue bar shows the mean LE with DSI method, and the red bar shows the mean LE obtained with the FDI method. The error bar shows the standard error of the mean.

High-Density vs Low-Density EEG

It is also observed that the LE increased while the number of scalp electrodes decreased (figure 5). The 2 discordant cases mentioned before were excluded from this analysis, as well. In the patients with intracranial recordings, statistically significant differences were found between 32-channel vs 19-channel montages (p = 0.0492), 76-channel vs 19-channel montages (p = 0.0169), and 76-channel vs 32-channel montages (p = 0.0461). In the patients who underwent surgical resection, and in the subset of patients who became seizure-free after lobectomy, statistically significant differences were also found between 32-channel vs 19-channel montages (p = 0.0107) and 76-channel vs 19-channel montages (p = 0.0012). Details are available from Dryad (appendix, table S1): doi.org/10.5061/dryad.4b8gtht9n.

Figure 5. Overall Performance of Ictal Source Imaging.

Figure 5

Open in a new tab

Seventy-six– (Green), 32- (yellow), and 19- (red) electrodes montage were applied and compared in patients who underwent intracranial EEG (iEEG) monitoring (left panel), patients who had surgical resection (middle panel), and patients who had surgical lobectomy resection and became seizure-free (right panel). In the left panel, DSI results were compared to iEEG electrodes location. In the middle and right panels, DSI results were compared to the resected volume. Increasing the number of scalp electrodes significantly improves the localization accuracy when evaluated through both iEEG and surgical resection. The number of patients for each group is 18 for iEEG, 26 for resection, and 15 for seizure-free after surgical resection. LE = localization error.

Interictal EEG Analysis Outcome

For interictal study, at least 10 hours (or equivalent) of artifact-free recordings were included for each patient. A total of 1,325 IEDs were identified and 59 clusters were formed from these IEDs (1.52 clusters per patient, SD 0.82). Distribution of IED cluster numbers for all patients is shown in figure 6A. When the dominant IED cluster is used for source imaging, namely the major cluster, the concordance rate is 78.4%. Selecting the dominant cluster, i.e., cluster with maximum number of IEDs, is the most reasonable strategy when a patient demonstrates multiple IED types, a priori. The closest cluster can only be defined retrospectively is are provided here to show the lower limit for LE.

Figure 6. Summary of Interictal Analysis Results and Comparison to Ictal Analysis Results.

Figure 6

Open in a new tab

(A) The distribution of interictal epileptiform discharge (IED) cluster numbers for each patient. We found no IED in 2 patients and 1 dominant IED cluster in 21 patients. In 16 patients, more than 1 IED cluster was identified. (B) The distribution of outcomes grouped by the number of IED clusters. For patients with only 1 IED cluster, the majority were seizure-free (International League Against Epilepsy [ILAE] 1–2); for patients with more than 1 IED cluster, the majority were not seizure-free (ILAE 3–6). (C) The sublobar concordance rate from ictal source imaging results. (D) The sublobar concordance rate from interictal source imaging results. The IED cluster with dominant occurrence was used for sublobar concordance evaluation. Two patients with no IED found were excluded. (E) Localization error (LE) (in mm) comparison between ictal source imaging and interictal source imaging in 18 patients with intracranial recordings. (F) LE (in mm) comparison between ictal source imaging and interictal source imaging in patients with surgical resections. DSI = dynamic seizure imaging.

For 18 patients who underwent iEEG recording and have at least 1 IED cluster, the mean LE for all clusters is about 3.67 cm (SD 2.77 cm, min 0.53 cm, max 12.8 cm). However, if the dominant IED cluster is used to calculate the localization error, the mean LE is about 3.12 cm (SD 2.01 cm, min 0.53 cm, max 7.18 cm). Using the closest cluster (with smallest LE among a patient's clusters) to evaluate the error gives an LE of 2.22 cm (SD 1.43 cm, min 0.53 cm, max 5.44 cm), which is still substantially higher than the ictal source imaging LEs. For 27 patients who had postoperative MRI and had at least 1 IED cluster, the mean LE for all clusters is about 1.42 cm (SD 1.82 cm, min 0 cm, max 6.71 cm).

Interictal EEG vs Ictal EEG

Qualitative comparisons between ictal and interictal analysis results are shown in figure 6, C and D. The sublobar concordance rate of ictal source imaging is 94.87% (37/39), which is better than interictal source imaging. A more significant difference is seen in patients with more than 1 IED cluster, where the ictal source imaging shows a concordance rate of 87.5%, while interictal source imaging shows a concordance rate of 56.2%.

Quantitative comparisons are shown in figure 6, E and F. The 2 discordant cases, mentioned before, were excluded for the qualitative analysis. In a group of 18 patients who underwent iEEG recordings, the LEs of ictal source imaging were compared with the LEs of interictal source imaging in the 3 cluster groups (all clusters, major cluster, and closest cluster). The LEs of ictal source imaging are significantly smaller than LE of all clusters (p = 0.0005) and major cluster (p = 0.0078). In a subset of patients who were not seizure-free, the differences are also significant (p = 0.0027) when comparing to all clusters and when comparing to the major cluster (p = 0.0318). In the subset of 26 patients whose surgical resection information was available, the LEs of ictal source imaging are not significantly different from interictal source imaging results.

Discussion

The present study demonstrates the clinical applicability and merits of a noninvasive electrophysiologic seizure imaging approach, which is currently missing in clinical routine, but may become an important component of the presurgical evaluation routine for epilepsy treatment. All the noninvasive ictal analysis results, except 2 with negative surgical outcome, are sublobar concordant with the gold standard of invasive clinical evaluations (iEEG or surgical resection). In our study, we observed that in a group of 39 patients with partial seizures, the proposed 76-channel dense-array EEG surpassed the conventional low-density EEG systems by its superior performance and accuracy. In general, the high degree of concordance between the ictal source imaging results and clinical findings was associated with an increased likelihood of a seizure-free postoperative outcome. This alludes to the accuracy of our proposed approach despite using noninvasive EEG recordings and attests to the possibility of applying this approach to help guide surgical planning or iEEG electrode placement.

Several other studies on EEG ictal source imaging have been reported in recent years. Few researchers13,15,28,29 studied patients with epilepsy with low-density EEG, where qualitative metrics, such as hemisphere/lobe-wise concordance, were reported to assess the outcome. Magnetoencephalography ictal source imaging results were also reported14,30 in some patients; however, the patients participating in such studies are typically constrained by their seizure types and rate. To our knowledge, for the purpose of clinical application, most of the studies on ictal source localization were limited by the relatively small number of patients or low-density EEG electrodes, mainly due to the difficulties in obtaining high-density seizure recordings. Moreover, the values of ictal recording over conventional interictal recording were often overlooked, as very few of these studies compared ictal source imaging and interictal source imaging results in the same group of patients with clinical ground truth such as iEEG or surgical resection volume. In our recent publication31 focusing on the precise estimation of source extent, ictal source imaging showed better performance over interictal activities. However, the IEDs were not processed to reflect clinical practice (i.e., the comprehensive clustering approach adopted in this work) and the current dynamic seizure source imaging algorithm provides a computationally more efficient approach than the extent estimation approach.29 A comprehensive study is needed to validate the merit of ictal source imaging in terms of its quantitative performance, and to address the key issues during the clinical translation process. The present work provides comprehensive investigation of the quantitative source imaging accuracy and comparison between ictal and interictal source imaging in a relatively large patient group with rigorous comparison to clinical ground truth as determined from iEEG and surgical resection volume. Such a study is important to establish clinical EEG source imaging for aiding surgical planning in patients with epilepsy.

Invasive iEEG monitoring is the current clinical gold standard for determining the SOZ.32 However, the risks and complications associated with invasive approaches restrict the number of patients who can benefit from such a modality. The present study suggests a noninvasive approach that provides spatiotemporal imaging in the same spatial domain as iEEG with the added benefit that EEG electrodes have a broader spatial coverage. In the cohort of patients studied here, our noninvasive approach localizes the SOZ with a 1.35 cm LE from SOZ electrodes determined from iEEG. Given that the LEs obtained by our noninvasive imaging approach are close to 1.3 cm, which is close to the iEEG interelectrode distance, the accuracy of our proposed algorithm in localizing ictal activity is approximately approaching iEEG resolution (1 cm), which can potentially have up to 1 cm error in determining foci of activity. The close proximity of the noninvasive imaging results to iEEG invasive measurements suggests the value of our dense-array EEG-based imaging approach in localizing ictal sources, possibly supplanting invasive procedures in some patients. Our noninvasive approach that has whole-brain coverage can guide iEEG electrode placement to improve its accuracy and diagnostic yield.

A sufficient number of EEG recording channels ensures needed spatial sampling at scalp.33 In this study, we evaluated the seizure imaging results using 76, 32, and 19 electrodes in a relatively large group of patients. We showed that in the patients with intracranial recordings, localization accuracy with 76-channel recordings was significantly better than 32-channel as well as 19-channel, indicating the merits of using higher density of EEG recordings as commonly practiced in most clinical settings. While the difference was not significant between 32-channel and 76-channel recordings, when comparing with resection, it is also widely acknowledged that the resected region may include more brain volume than identified SOZ to ensure seizure-free outcomes. Thus, while higher density EEG recordings may not have a significant effect on the surgical outcomes of resection, it may influence the outcome of neuromodulation or laser ablation procedures that attempt to treat focal epilepsy by injecting physical energy to epileptic foci. Such procedures need to determine the epileptogenic foci to ultimately determine the target of stimulation/ablation. Despite the significant improvement of source localization performance, the cost of using the 76-channel system compared to the lower-density EEG recording systems is the clinical burden of only 30 more minutes of preparation time to mount the additional electrodes and, possibly, to reattach and maintain electrodes during the recording. Considering the benefits of having such a noninvasive and economical, but high-resolution electrophysiologic imaging approach, and the fact that clinical EEG monitoring is usually long-term, the increased 30-minute preparation time is minimal.

Our study demonstrates a significantly higher accuracy of localizing SOZ from ictal source imaging than interictal source imaging. The clear evidence that seizures, compared to IEDs, more reliably determine the epileptogenic zone is well-stated in previous studies34–36 as well as ours. Despite the technical challenge of recording seizures in preoperative settings compared with IED recordings, the discrepancy between regions of spiking and SOZ has been discussed.37–39 In a study involving 271 patients with unilateral temporal lobe epilepsy, Vollmar et al.18 found that 38.8% of the patients had discordant interictal activities. In a study comparing automatic spike detection with clinician-aided spike detection, Baroumand et al.40 also found that spike clusters with the most spike numbers could be different from clinician's diagnosis and choice of spike. In our study, particularly in patients with worse outcomes (ILAE 3–6), more complicated interictal activities are identified (60% have more than 1 IED cluster) and the dominant IED clusters in these patients show a clear deviation from intracranial results, which is also observed in previous studies.41 This indicates that multiple inconsistent spike clusters result in inconsistent estimates that will not match iEEG SOZ findings, and it is important to resolve such inconsistencies with ictal imaging. Our ictal imaging estimates, in the same patients, obtained more consistent results. Even though these patients were not completely seizure-free after surgical resection, our noninvasive ictal source imaging estimations are concordant with clinical decisions made by epilepsy experts who have various information inputs, including invasive recordings. A potential hypothesis suggested by a few iEEG studies42,43 is that resecting both SOZ and IZ may increase surgical success, although such findings barely confound the importance of localizing SOZ as they still consider SOZ as the primary target in surgical intervention. We further analyzed the effect of ictal onset patterns as literature suggested44,45 on our proposed algorithm's performance and found some marginal effects, but such investigations deserve their own separate study (data available from Dryad, appendix, doi.org/10.5061/dryad.4b8gtht9n).

In a large group of patients with surgical outcome and follow-ups, our analysis showed higher accuracy in patients with ILAE 1–2 outcome compared to patients with ILAE 3–6 outcome. For some of the ILAE 3–6 patients, the reconstructed source activations were not covered by the iEEG electrodes or were not included in the resection, which may suggest potential epileptogenic foci that were missed and ultimately were not surgically removed. Nonetheless, the reason for such surgical outcome is not yet clear and as no further surgeries were conducted on those patients so far, our aforementioned interpretation cannot be validated directly. With recent development of extensive multiscale iEEG electrode implantation and brain stimulation protocols,46,47 the connectivity among different brain regions can be probed more accurately to ultimately identify the epileptogenic onset.

Although the present approach has notable merits, its clinical application is not without limitations. Among different types of patients with epilepsy, we only tested the method on patients with focal epilepsy where, initially, only one hemisphere is involved during an ictal event. While the method itself would be applicable to any kind of electrophysiologic activity, we chose to study focal epilepsy as it is possible to obtain a clear SOZ, or a ground truth, to validate our results. Thus, if the method is to be applied on the EEG of a patient with generalized epilepsy, the results should be interpreted with caution. Furthermore, among the 28 operated patients, 19 patients have temporal epilepsy, which reflects the common focal epilepsy patient population. Although the ratio is already low among studies with comparable number of patients, studies involving more patients with extratemporal lobe epilepsy would be desired for further validation. Moreover, localizing deep brain activities from scalp EEG is usually more challenging and electrical source imaging performance might decrease due to lower SNR of generated signals recorded at scalp compared to superficial source localization. Although a few patients with deep foci were involved in the study and comparable performance was observed for them, our results indicate that deep brain activation needs to be carefully interpreted in practice. As our data showed, the accuracy of localizing SOZ decreased significantly as the number of EEG electrodes decreased, which indicates that our proposed method works optimally when combined with high-density EEG recording systems (76-channel or more). Our proposed method can be adapted to low-density recordings; however, the results have to be interpreted more carefully as the accuracy most likely will be affected. Moreover, in our proposed approach, high-quality MRIs were used to construct individual boundary element head models for each patient, as MRI is usually part of the presurgical evaluation for patients with epilepsy. In rare cases where MRI is not available, generic models or spherical models can be an alternative,11 although the localization accuracy would be decreased.

Despite these limitations, high-resolution EEG monitoring in combination with the DSI approach provides a framework to noninvasively assess the rapid electrophysiologic abnormalities during seizures. The concordance of the imaging results with iEEG and surgical outcome suggests a potential application for such a protocol in routine presurgical evaluation to (1) aid the implantation of iEEG electrodes in order to improve the localization yield and accuracy of iEEG and in some cases to (2) assist the surgical resection planning as an alternative for the invasive long-term iEEG monitoring. The development of such a high-density EEG recording and DSI method would meet the need for a noninvasive electrophysiologic seizure imaging protocol, currently missing in the clinical routine.

Acknowledgment

The authors thank the students and personnel of the Biomedical Functional Imaging and Neuroengineering Lab and the Mayo Systems Electrophysiology Lab for helpful discussions.

Glossary

DSI

dynamic seizure imaging

EZ

epileptogenic zone

FDI

frequency domain–based imaging

IED

interictal epileptiform discharge

iEEG

intracranial EEG

ILAE

International League Against Epilepsy

IZ

irritative zone

LE

localization error

SNR

signal-to-noise ratio

SOZ

seizure onset zone

Appendix. Authors

Appendix.

Open in a new tab

Study Funding

Supported by NIH EB021027, NS096761, EB006433, and EB007920.

Disclosure

The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.

References

  • 1.Ryvlin P, Cross JH, Rheims S. Epilepsy surgery in children and adults. Lancet Neurol 2014;13:1114–1126. [DOI] [PubMed] [Google Scholar]
  • 2.Duncan JS, Winston GP, Koepp MJ, Ourselin S. Brain imaging in the assessment for epilepsy surgery. Lancet Neurol 2016;15:420–433. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Rosenow F, Lüders H. Presurgical evaluation of epilepsy. Brain 2001;124:1683–1700. [DOI] [PubMed] [Google Scholar]
  • 4.Dwivedi R, Ramanujam B, Chandra PS, et al. Surgery for drug-resistant epilepsy in children. N Engl J Med 2017;377:1639–1647. [DOI] [PubMed] [Google Scholar]
  • 5.Duncan JS. Imaging in the surgical treatment of epilepsy. Nat Rev Neurol 2010;6:537–550. [DOI] [PubMed] [Google Scholar]
  • 6.Noe K, Sulc V, Wong-Kisiel L, et al. Long-term outcomes after nonlesional extratemporal lobe epilepsy surgery. JAMA Neurol 2013;70:1003–1008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Blount JP, Cormier J, Kim H, Kankirawatana P, Riley KO, Knowlton RC. Advances in intracranial monitoring. Neurosurg Focus 2008;25:E18. [DOI] [PubMed] [Google Scholar]
  • 8.Brodbeck V, Spinelli L, Lascano AM, et al. Electroencephalographic source imaging: a prospective study of 152 operated epileptic patients. Brain 2011;134:2887–2897. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.He B, Sohrabpour A, Brown E, Liu Z. Electrophysiological source imaging: a noninvasive window to brain dynamics. Annu Rev Biomed Eng 2018;20:171–196. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Michel CM, Bin H. EEG mapping and source imaging. In: Schomer DL, Lopes da Silva FH, eds. Niedermeyer's Electroencephalography: Basic Principles, Clinical Applications, and Related Fields, 7th ed. Philadelphia: Lippincott Williams & Wilkins; 2017:1135–1156. [Google Scholar]
  • 11.Kaiboriboon K, Lüders HO, Hamaneh M, Turnbull J, Lhatoo SD. EEG source imaging in epilepsy: practicalities and pitfalls. Nat Rev Neurol 2012;8:498–507. [DOI] [PubMed] [Google Scholar]
  • 12.Knowlton RC, Elgavish RA, Bartolucci A, et al. Functional imaging: II: prediction of epilepsy surgery outcome. Ann Neurol 2008;64:35–41. [DOI] [PubMed] [Google Scholar]
  • 13.Nemtsas P, Birot G, Pittau F, et al. Source localization of ictal epileptic activity based on high-density scalp EEG data. Epilepsia 2017;58:1027–1036. [DOI] [PubMed] [Google Scholar]
  • 14.Pellegrino G, Hedrich T, Chowdhury R, et al. Source localization of the seizure onset zone from ictal EEG/MEG data. Hum Brain Mapp 2016;37:2528–2546. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Beniczky S, Lantz G, Rosenzweig I, et al. Source localization of rhythmic ictal EEG activity: a study of diagnostic accuracy following STARD criteria. Epilepsia 2013;54:1743–1752. [DOI] [PubMed] [Google Scholar]
  • 16.Assaf BA, Karkar KM, Laxer KD, et al. Ictal magnetoencephalography in temporal and extratemporal lobe epilepsy. Epilepsia 2003;44:1320–1327. [DOI] [PubMed] [Google Scholar]
  • 17.Alarcon G, Guy CN, Binnie CD, Walker SR, Elwes RD, Polkey CE. Intracerebral propagation of interictal activity in partial epilepsy: implications for source localisation. J Neurol Neurosurg Psychiatry 1994;57:435–449. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Vollmar C, Stredl I, Heinig M, Noachtar S, Rémi J. Unilateral temporal interictal epileptiform discharges correctly predict the epileptogenic zone in lesional temporal lobe epilepsy. Epilepsia 2018;59:1577–1582. [DOI] [PubMed] [Google Scholar]
  • 19.Rémi J, Vollmar C, De Marinis A, Heinlin J, Peraud A, Noachtar S. Congruence and discrepancy of interictal and ictal EEG with MRI lesions in focal epilepsies. Neurology 2011;77:1383–1390. [DOI] [PubMed] [Google Scholar]
  • 20.Ergene E, Shih JJ, Blum DE, So NK. Frequency of bitemporal independent interictal epileptiform discharges in temporal lobe epilepsy. Epilepsia 2000;41:213–218. [DOI] [PubMed] [Google Scholar]
  • 21.Fisher RS, Van Emde Boas W, Blume W, et al. Epileptic seizures and epilepsy: definitions proposed by the international league against epilepsy (ILAE) and the international bureau for epilepsy (IBE). Epilepsia 2005;46:470–472. [DOI] [PubMed] [Google Scholar]
  • 22.Yang L, Wilke C, Brinkmann B, Worrell GA, He B. Dynamic imaging of ictal oscillations using non-invasive high-resolution EEG. Neuroimage 2011;56:1908–1917. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Yuan H, Doud A, Gururajan A, He B. Cortical imaging of event-related (de)synchronization during online control of brain-computer interface using minimum-norm estimates in frequency domain. IEEE Trans Neural Syst Rehabil Eng 2008;16:425–431. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Yuan H, Liu T, Szarkowski R, Rios C, Ashe J, He B. Negative covariation between task-related responses in alpha/beta-band activity and BOLD in human sensorimotor cortex: an EEG and fMRI study of motor imagery and movements. Neuroimage 2010;49:2596–2606. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Pascual-Marqui RD. Standardized low resolution brain electromagnetic tomography (sLORETA): technical details. Methods Find Exp Clin Pharmacol 2002;24:5–12. [PubMed] [Google Scholar]
  • 26.Agirre-Arrizubieta Z, Huiskamp GJM, Ferrier CH, Van Huffelen AC, Leijten FSS. Interictal magnetoencephalography and the irritative zone in the electrocorticogram. Brain 2009;132:3060–3071. [DOI] [PubMed] [Google Scholar]
  • 27.Huiskamp G, Agirre-Arrizubieta Z, Leijten F. Regional differences in the sensitivity of MEG for interictal spikes in epilepsy. Brain Topogr 2010;23:159–164. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Lantz G, Michel CM, Seeck M, et al. Space-oriented segmentation and 3-dimensional source reconstruction of ictal EEG patterns. Clin Neurophysiol 2001;112:688–697. [DOI] [PubMed] [Google Scholar]
  • 29.Staljanssens W, Strobbe G, Holen RV, et al. Seizure onset zone localization from ictal high-density EEG in refractory focal epilepsy. Brain Topogr 2017;30:257–271. [DOI] [PubMed] [Google Scholar]
  • 30.Alkawadri R, Burgess RC, Kakisaka Y, Mosher JC, Alexopoulos AV. Assessment of the utility of ictal magnetoencephalography in the localization of the epileptic seizure onset zone. JAMA Neurol 2018;75:1264–1272. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Sohrabpour A, Cai Z, Ye S, Brinkmann B, Worrell G, He B. Noninvasive electromagnetic source imaging of spatiotemporally distributed epileptogenic brain sources. Nat Commun 2020;11:1946. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Jayakar P, Gotman J, Harvey AS, et al. Diagnostic utility of invasive EEG for epilepsy surgery: indications, modalities, and techniques. Epilepsia 2016;57:1735–1747. [DOI] [PubMed] [Google Scholar]
  • 33.Lantz G, de Peralta RG, Spinelli L, Seeck M, Michel CM. Epileptic source localization with high density EEG: how many electrodes are needed? Clin Neurophysiol 2003;114:63–69. [DOI] [PubMed] [Google Scholar]
  • 34.Karoly PJ, Freestone DR, Boston R, et al. Interictal spikes and epileptic seizures: their relationship and underlying rhythmicity. Brain 2016;139:1066–1078. [DOI] [PubMed] [Google Scholar]
  • 35.Staley K, Hellier JL, Dudek FE. Do interictal spikes drive epileptogenesis? Neuroscientist 2005;11:272–276. [DOI] [PubMed] [Google Scholar]
  • 36.Habib MA, Ibrahim F, Mohktar MS, Kamaruzzaman SB, Rahmat K, Lim KS. Ictal EEG source imaging for presurgical evaluation of refractory focal epilepsy. World Neurosurg 2016;88:576–585. [DOI] [PubMed] [Google Scholar]
  • 37.Marsh ED, Peltzer B, Brown MW, et al. Interictal EEG spikes identify the region of electrographic seizure onset in some, but not all, pediatric epilepsy patients. Epilepsia 2010;51:592–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Alarcon G. Electrophysiological aspects of interictal and ictal activity in human partial epilepsy. Seizure 1996;5:7–33. [DOI] [PubMed] [Google Scholar]
  • 39.Hufnagel A, Dümpelmann M, Zentner J, Schijns O, Elger CE. Clinical relevance of quantified intracranial interictal spike activity in presurgical evaluation of epilepsy. Epilepsia 2000;41:467–478. [DOI] [PubMed] [Google Scholar]
  • 40.Baroumand AG, van Mierlo P, Strobbe G, et al. Automated EEG source imaging: a retrospective, blinded clinical validation study. Clin Neurophysiol 2018;129:2403–2410. [DOI] [PubMed] [Google Scholar]
  • 41.Bartolomei F, Trébuchon A, Bonini F, et al. What is the concordance between the seizure onset zone and the irritative zone? A SEEG quantified study. Clin Neurophysiol Int Fed Clin Neurophysiol 2016;127:1157–1162. [DOI] [PubMed] [Google Scholar]
  • 42.Krsek P, Maton B, Jayakar P, et al. Incomplete resection of focal cortical dysplasia is the main predictor of poor postsurgical outcome. Neurology 2009;72:217–223. [DOI] [PubMed] [Google Scholar]
  • 43.Wyllie E, Luders H, Morris HH III, et al. Clinical outcome after complete or partial cortical resection for intractable epilepsy. Neurology 1987;37:1634–1641. [DOI] [PubMed] [Google Scholar]
  • 44.Perucca P, Dubeau F, Gotman J. Intracranial electroencephalographic seizure-onset patterns: effect of underlying pathology. Brain 2014;137:183–196. [DOI] [PubMed] [Google Scholar]
  • 45.Tanaka H, Khoo HM, Dubeau F, Gotman J. Association between scalp and intracerebral electroencephalographic seizure-onset patterns: a study in different lesional pathological substrates. Epilepsia 2018;59:420–430. [DOI] [PubMed] [Google Scholar]
  • 46.Stead M, Bower M, Brinkmann BH, et al. Microseizures and the spatiotemporal scales of human partial epilepsy. Brain 2010;133:2789–2797. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Grewal SS, Benscoter M, Kuehn S, et al. Minimally invasive, endoscopic-assisted device for subdural electrode implantation in epilepsy. Oper Neurosurg 2020;8:92–97. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Articles from Neurology are provided here courtesy of American Academy of Neurology

RESOURCES